matplotlib-devel Mailing List for matplotlib (Page 3)

On 04/18/06 18:04, Charlie Moad wrote:
> Thanks for the hard work, Ken, but I think I officially claim defeat
> on this. After using your patch I got a few obvious errors then ran
> into this (pasted below).
Sweet, sweet INTERNAL COMPILER ERROR. I think the Boost folks have a strategy 
for working around this situation by hoisting the friend function template 
entirely outside of the class. I'll see what I can find online and try to 
submit another patch soon.
Ken

On 4/13/06, Ken McIvor <mc...@ii...> wrote:
> On Apr 13, 2006, at 9:23 PM, Charlie Moad wrote:
> >> I'm thoroughly confused about how the STL comes into play if VC++6 is
> >> dying because it can't handle the code constructs in use.
> >
> > My impression of the error is that the first argument, which is a
> > template, is confusing the vs6 compiler. There are docs everywhere
> > saying that vs6 doesn't handle templates worth a darn. All this
> > together forms my hunch, but I may be totally wrong.
>
> Templates are a feature of the C++ language. It sounds like you're
> running up against fundamental limitations of the compiler. If that
> is the case, then changing from one implementation of the STL to
> another probably won't do anything.
>
> >> I'm not following John's comment about the animation blit... I
> >> thought that the
> >> _n?_backend_agg module(s) are part of the Agg backend. Is John
> >> proposing we drop Agg support under Python 2.3 / VC++6?
> >
> > So the reason that I made the vs7 build for python2.4 is that wx
> > blitting in windows was not supported. Mingw is unable to link
> > against the libs distributed by wxpython since they are created with
> > vs. All John is saying is that people can probably live without wx
> > blitting in python2.3 (we would still distribute the mingw 2.3 build).
> > The tk and gtk aggs can still blit fine.
>
> Ah, I think I understand now. The WxAgg backend supports blitting
> without the _wxagg module, but it won't be as fast.
>
> > Sounds great. I'll be more than happy to test out any of your
> > changes. I am leaving town for the weekend tomorrow for Easter, so
> > there may be a delay.
>
> Attached is a patch against svn 2292. It works with GGC 3.3.6 on
> Debian Stable. I'm not confident it will work under windows... as I
> was wrapping it up I found that boost.org claims that "a template
> cannot be declared a friend of a class" [1]. I guess it's worth a
> shot, though.
>
> Ken
>
> [1] http://www.boost.org/more/microsoft_vcpp.html
>
Thanks for the hard work, Ken, but I think I officially claim defeat
on this. After using your patch I got a few obvious errors then ran
into this (pasted below).
Thanks again,
 Charlie
building tkagg 2 3
 Building for python23
running build
running build_py
running build_ext
No module named msvccompiler in numpy.distutils, trying from distutils..
building 'matplotlib.backends._na_backend_agg' extension
C:\Program Files\Microsoft Visual Studio\VC98\BIN\cl.exe /c /nologo /Ox /MD=
 /W3
/GX /DNDEBUG -Iwin32_static\include -I. -Isrc -Iswig -Iagg23/include -I. -I=
win32
_static\include -I. -Iwin32_static\include\freetype2 -I.\freetype2 -Isrc\fr=
eetyp
e2 -Iswig\freetype2 -Iagg23/include\freetype2 -I.\freetype2 -Iwin32_static\=
inclu
de\freetype2 -I.\freetype2 -Ic:\Python23\include -Ic:\Python23\PC /Tpsrc/_n=
a_bac
kend_agg.cpp /Fobuild\temp.win32-2.3\Release\src/_na_backend_agg.obj -DNUMA=
RRAY=3D
1
_na_backend_agg.cpp
src/_na_backend_agg.cpp(63) : warning C4800: 'long' : forcing value to bool=
 'tru
e' or 'false' (performance warning)
C:\Program Files\Microsoft Visual Studio\VC98\INCLUDE\xlocnum(267) : fatal =
error
 C1001: INTERNAL COMPILER ERROR
 (compiler file 'msc1.cpp', line 1794)
 Please choose the Technical Support command on the Visual C++
 Help menu, or open the Technical Support help file for more inform=
ation
 C:\Program Files\Microsoft Visual Studio\VC98\INCLUDE\xlocnum(263) =
: whi
le compiling class-template member function 'class std::istreambuf_iterator=
<char
,struct std::char_traits<char> > __thiscall std::num_get<char,class std::is=
tream
buf_iterator<char,struct std::char_traits<char> > >::do_get(class std::istr=
eambu
f_iterator<char,struct std::char_traits<char> >,class std::istreambuf_itera=
tor<c
har,struct std::char_traits<char> >,class std::ios_base &,int &,void *& ) c=
onst'
error: Command ""C:\Program Files\Microsoft Visual Studio\VC98\BIN\cl.exe" =
/c /n
ologo /Ox /MD /W3 /GX /DNDEBUG -Iwin32_static\include -I. -Isrc -Iswig -Iag=
g23/i
nclude -I. -Iwin32_static\include -I. -Iwin32_static\include\freetype2 -I.\=
freet
ype2 -Isrc\freetype2 -Iswig\freetype2 -Iagg23/include\freetype2 -I.\freetyp=
e2 -I
win32_static\include\freetype2 -I.\freetype2 -Ic:\Python23\include -Ic:\Pyt=
hon23
\PC /Tpsrc/_na_backend_agg.cpp /Fobuild\temp.win32-2.3\Release\src/_na_back=
end_a
gg.obj -DNUMARRAY=3D1" failed with exit status 2

>>>>> "Gary" == Gary Ruben <gr...@bi...> writes:
 Gary> Since noone has picked this up and I notice John Hunter is
 Gary> active at the moment, I'll ask directly whether you, John,
 Gary> can apply my figimage patches. Better now than when the
 Gary> diffed versions become outdated (some may be
 Gary> already). Alternatively, please let me know if I should be
 Gary> posting it somewhere so it isn't forgotten. thanks, Gary R.
Thanks for the prompting -- I committed some changes that should fix
this bug -- give svn revision 2301 a try.
JDH

Hello everyone,
I have the following problem with my application, which allows you to 
manipulate some of properties of a plot through the use of a gui-style 
property editor:
Whenever I set the y-axis scale to logarithmic, but have the y range 
include a negative value, I will get a ValueError upon expose_event. The 
bad thing is now, that my application catches all exceptions and 
displays a user-friendly dialog, explaining or listing the exception and 
offering the user to mail the output to the dumb author of the program. 
But whenever the user quits the dialog, a new expose event of the canvas 
will be triggered, leading to an unstoppable series of exceptions 
(except when using xkill of course).
So what I would like to propose is a modification of expose_event 
located in backends/backend_gtk.py to better handle any exceptions that 
it might trigger.
One not so nice solution of mine was to wrap the method in a 
try...finally statement, which ensures that False is returned and the 
expose_event is stopped:
 def expose_event(self, widget, event):
 """Expose_event for all GTK backends. Should not be overridden.
 """
 if _debug: print 'FigureCanvasGTK.%s' % fn_name()
 try:
 if GTK_WIDGET_DRAWABLE(self):
 if self._need_redraw:
 x, y, w, h = self.allocation
 self._pixmap_prepare (w, h)
 self._render_figure(self._pixmap, w, h)
 self._need_redraw = False
 x, y, w, h = event.area
 self.window.draw_drawable (self.style.fg_gc[self.state],
 self._pixmap, x, y, x, y, w, h)
 finally:
 return False # finish event propagation?
Of course, this has the obvious disadvantage, that no exception is 
triggered at all, which is undesirable.
Are there any better approaches to this?
Best regards,
Niklas Volbers.

Since noone has picked this up and I notice John Hunter is active at the 
moment, I'll ask directly whether you, John, can apply my figimage 
patches. Better now than when the diffed versions become outdated (some 
may be already). Alternatively, please let me know if I should be 
posting it somewhere so it isn't forgotten.
thanks,
Gary R.

On Tuesday 18 April 2006 10:45, John Hunter wrote:
> >>>>> "Darren" == Darren Dale <dd...@co...> writes:
>
> Darren> This is fixed in svn-2298.
>
> OK, thanks. It looks like the superscript formatting is wrong in the
> offset text in usetex=False (eg see the '10' superscript in the
> 10^{-10} in subplot(221) in figure(4)).
>
> and look at the string below the xaxis of subplot(221) (the one that
> reads x10^5 + 1x10^10. The 10^{10} part looks like
>
> {1^0}^{1^0}
>
> on my system. This could be a mathtext layout bug, or maybe the
> syntax of the string in the offset text is wrong. If you can confirm
> that the latex is correct, could you post the string and I'll see if I
> can debug the layout problem?
I am sorry to report that it's a mathtext layout bug.
Darren

>>>>> "Darren" == Darren Dale <dd...@co...> writes:
 Darren> This is fixed in svn-2298.
OK, thanks. It looks like the superscript formatting is wrong in the
offset text in usetex=False (eg see the '10' superscript in the
10^{-10} in subplot(221) in figure(4)).
and look at the string below the xaxis of subplot(221) (the one that
reads x10^5 + 1x10^10. The 10^{10} part looks like
 {1^0}^{1^0}
on my system. This could be a mathtext layout bug, or maybe the
syntax of the string in the offset text is wrong. If you can confirm
that the latex is correct, could you post the string and I'll see if I
can debug the layout problem?
JDH

On Tuesday 18 April 2006 10:08, John Hunter wrote:
> The axis offset text does not work if usetex=False (eg it has a
> "\times" string in it). Darren, could you take a look at this when
> you get a minute and see if it can be made to work with usetex =
> True|False.
>
> When I try and run the newscalarformatter_demo.py example
>
>
> Traceback (most recent call last):
> File
> "/usr/lib/python2.4/site-packages/matplotlib/backends/backend_gtk.py",
> line 293, in expose_event
> self._render_figure(self._pixmap, w, h)
> File
> "/usr/lib/python2.4/site-packages/matplotlib/backends/backend_gtkagg.py",
> line 72, in _render_figure
> FigureCanvasAgg.draw(self)
> File
> "/usr/lib/python2.4/site-packages/matplotlib/backends/backend_agg.py",
> line 398, in draw
> self.figure.draw(renderer)
> File "/usr/lib/python2.4/site-packages/matplotlib/figure.py", line
> 529, in draw
> for a in self.axes: a.draw(renderer)
> File "/usr/lib/python2.4/site-packages/matplotlib/axes.py", line
> 1677, in draw
> self.xaxis.draw(renderer)
> File "/usr/lib/python2.4/site-packages/matplotlib/axis.py", line
> 600, in draw
> self.offsetText.draw(renderer)
> File "/usr/lib/python2.4/site-packages/matplotlib/text.py", line
> 339, in draw
> bbox, info = self._get_layout(renderer)
> File "/usr/lib/python2.4/site-packages/matplotlib/text.py", line
> 186, in _get_layout
> w,h = renderer.get_text_width_height(
> File
> "/usr/lib/python2.4/site-packages/matplotlib/backends/backend_agg.py",
> line 253, in get_text_width_height
> width, height, fonts = math_parse_s_ft2font(
> File "/usr/lib/python2.4/site-packages/matplotlib/mathtext.py", line
> 1340, in math_parse_s_ft2font
> handler.expr.set_size_info(fontsize, dpi)
> File "/usr/lib/python2.4/site-packages/matplotlib/mathtext.py", line
> 976, in set_size_info
> self.elements[0].set_size_info(self._scale*fontsize, dpi)
> File "/usr/lib/python2.4/site-packages/matplotlib/mathtext.py", line
> 888, in set_size_info
> self.metrics = Element.fonts.get_metrics(
> File "/usr/lib/python2.4/site-packages/matplotlib/mathtext.py", line
> 402, in get_metrics
> cmfont, metrics, glyph, offset = \
> File "/usr/lib/python2.4/site-packages/matplotlib/mathtext.py", line
> 422, in _get_info
> raise ValueError('unrecognized symbol "%s"' % sym)
> ValueError: unrecognized symbol "\times10"
This is fixed in svn-2298.
Darren

The axis offset text does not work if usetex=False (eg it has a
"\times" string in it). Darren, could you take a look at this when
you get a minute and see if it can be made to work with usetex =
True|False.
When I try and run the newscalarformatter_demo.py example
Traceback (most recent call last):
 File
 "/usr/lib/python2.4/site-packages/matplotlib/backends/backend_gtk.py",
 line 293, in expose_event
 self._render_figure(self._pixmap, w, h)
 File
 "/usr/lib/python2.4/site-packages/matplotlib/backends/backend_gtkagg.py",
 line 72, in _render_figure
 FigureCanvasAgg.draw(self)
 File
 "/usr/lib/python2.4/site-packages/matplotlib/backends/backend_agg.py",
 line 398, in draw
 self.figure.draw(renderer)
 File "/usr/lib/python2.4/site-packages/matplotlib/figure.py", line
 529, in draw
 for a in self.axes: a.draw(renderer)
 File "/usr/lib/python2.4/site-packages/matplotlib/axes.py", line
 1677, in draw
 self.xaxis.draw(renderer)
 File "/usr/lib/python2.4/site-packages/matplotlib/axis.py", line
 600, in draw
 self.offsetText.draw(renderer)
 File "/usr/lib/python2.4/site-packages/matplotlib/text.py", line
 339, in draw
 bbox, info = self._get_layout(renderer)
 File "/usr/lib/python2.4/site-packages/matplotlib/text.py", line
 186, in _get_layout
 w,h = renderer.get_text_width_height(
 File
 "/usr/lib/python2.4/site-packages/matplotlib/backends/backend_agg.py",
 line 253, in get_text_width_height
 width, height, fonts = math_parse_s_ft2font(
 File "/usr/lib/python2.4/site-packages/matplotlib/mathtext.py", line
 1340, in math_parse_s_ft2font
 handler.expr.set_size_info(fontsize, dpi)
 File "/usr/lib/python2.4/site-packages/matplotlib/mathtext.py", line
 976, in set_size_info
 self.elements[0].set_size_info(self._scale*fontsize, dpi)
 File "/usr/lib/python2.4/site-packages/matplotlib/mathtext.py", line
 888, in set_size_info
 self.metrics = Element.fonts.get_metrics(
 File "/usr/lib/python2.4/site-packages/matplotlib/mathtext.py", line
 402, in get_metrics
 cmfont, metrics, glyph, offset = \
 File "/usr/lib/python2.4/site-packages/matplotlib/mathtext.py", line
 422, in _get_info
 raise ValueError('unrecognized symbol "%s"' % sym)
ValueError: unrecognized symbol "\times10"
Thanks!
JDH

Jeff Whitaker wrote:
> Andrew Straw wrote:
>
>>
>> My recommendation is to move pyshapelib/lib/shapelib to lib/shapelib and
>> pyshapelib/lib/dbflib to lib/dbflib and keep a single setup.py file. 
>
> Hi Andrew: That sounds reasonable, go ahead and do it if you like.
Thanks, done.
> I'll get rid of that try/except. Turns out it's only Thuban that
> installs pyshapelib (AFAICT it doesn't exist as a standalone package).
OK, I temporarily turned it off, but perhaps you want to do something
more sophisticated.
Cheers!
Andrew

Andrew Straw wrote:
> Hi Jeff,
>
> I've been investigating what is required to get setuptools' "develop"
> command to work on the pyshapelib stuff in basemap.
>
> The "develop" command essentially creates a link from site-packages to
> the source directory (e.g. pyshapelib/lib) so that Python statements
> like "import shapelib" will load the file directly from your source
> tree. In other words, this takes the "python setup.py install" out of
> the module/package development cycle.
>
> Unfortunately, the implementation in setuptools seems unable to locate
> the package root when there are multiple packages at multiple roots
> specified using the "packages" and "package_dirs" arguments to a single
> distutils.core.setup() call. I think there are two alternative
> approaches if we want to get this to work, which I do.
>
> My recommendation is to move pyshapelib/lib/shapelib to lib/shapelib and
> pyshapelib/lib/dbflib to lib/dbflib and keep a single setup.py file.
>
> 
Hi Andrew: That sounds reasonable, go ahead and do it if you like.
> The other option is to make two separate setup.py files, one for
> "basemap" (which installs the matplotlib.toolkits.basemap python
> packages) and one for "pyshapelib" (which installs the shapelib and
> dbflib Python packages).
>
> I've tested both of these options, and they both work. Let me know if
> you want me to make the changes in svn.
>
> Then there's the issue of the "try: import shapelib except:" in the
> setup.py... I'm not in love with that style of testing, but I'm not sure
> why it's in there. Do other packages install shapelib and you don't want
> the possibility of overwriting them? Or is it just to save some build
> time? If its the former, wouldn't it be better to rename basemap's
> version so basemap doesn't get hit with an API change in an external
> package? If it's the latter, I think that's a false economy. Finally, we
> should at least qualify the except statement to only gobble and ImportError.
>
> Cheers!
> Andrew
> 
I'll get rid of that try/except. Turns out it's only Thuban that 
installs pyshapelib (AFAICT it doesn't exist as a standalone package). 
-Jeff
-- 
Jeffrey S. Whitaker Phone : (303)497-6313
NOAA/OAR/CDC R/CDC1 FAX : (303)497-6449
325 Broadway Web : http://www.cdc.noaa.gov/~jsw
Boulder, CO, USA 80305-3328 Office: Skaggs Research Cntr 1D-124

Hi Jeff,
I've been investigating what is required to get setuptools' "develop"
command to work on the pyshapelib stuff in basemap.
The "develop" command essentially creates a link from site-packages to
the source directory (e.g. pyshapelib/lib) so that Python statements
like "import shapelib" will load the file directly from your source
tree. In other words, this takes the "python setup.py install" out of
the module/package development cycle.
Unfortunately, the implementation in setuptools seems unable to locate
the package root when there are multiple packages at multiple roots
specified using the "packages" and "package_dirs" arguments to a single
distutils.core.setup() call. I think there are two alternative
approaches if we want to get this to work, which I do.
My recommendation is to move pyshapelib/lib/shapelib to lib/shapelib and
pyshapelib/lib/dbflib to lib/dbflib and keep a single setup.py file.
The other option is to make two separate setup.py files, one for
"basemap" (which installs the matplotlib.toolkits.basemap python
packages) and one for "pyshapelib" (which installs the shapelib and
dbflib Python packages).
I've tested both of these options, and they both work. Let me know if
you want me to make the changes in svn.
Then there's the issue of the "try: import shapelib except:" in the
setup.py... I'm not in love with that style of testing, but I'm not sure
why it's in there. Do other packages install shapelib and you don't want
the possibility of overwriting them? Or is it just to save some build
time? If its the former, wouldn't it be better to rename basemap's
version so basemap doesn't get hit with an API change in an external
package? If it's the latter, I think that's a false economy. Finally, we
should at least qualify the except statement to only gobble and ImportError.
Cheers!
Andrew

On 4/13/06, John Hunter <jdh...@ac...> wrote:
> >>>>> "Charlie" =3D=3D Charlie Moad <cw...@gm...> writes:
> Charlie> So this approach brought actually got me passed the
> Charlie> error, BUT it walked me straight into an internal
> Charlie> compiler error with just a code returned. I am thinking
> Charlie> I will have to start looking into these open source STL
> Charlie> implementations. I'll work on it more tonight and write
> Charlie> back with what I find out. Did I mention my progression
> Charlie> through major programming languages never included C++.
> Charlie> I do appreciate and really need the help.
>
> While I appreciate all the extra effort you are going to, we migfht
> take a step back and say that animation blit is not supported on
> python2.3 and win32. If you want the latest features, use the latest
> python. The wx backend and traditional animation will still work on
> python2.3. And with the enthought python2.4 release supposedly in the
> pipeline, I suspect most who need the faster techniques will be able
> to upgrade.
I guess I am also driven by not have to deal with mingw. The VS7
build is just so simple. There are a lot of sites claiming you can
just use the headers from STLport (stlport.org) and it just works. I
will try to get this working and may claim defeat if that fails.

>>>>> "Charlie" == Charlie Moad <cw...@gm...> writes:
 Charlie> So this approach brought actually got me passed the
 Charlie> error, BUT it walked me straight into an internal
 Charlie> compiler error with just a code returned. I am thinking
 Charlie> I will have to start looking into these open source STL
 Charlie> implementations. I'll work on it more tonight and write
 Charlie> back with what I find out. Did I mention my progression
 Charlie> through major programming languages never included C++.
 Charlie> I do appreciate and really need the help.
While I appreciate all the extra effort you are going to, we migfht
take a step back and say that animation blit is not supported on
python2.3 and win32. If you want the latest features, use the latest
python. The wx backend and traditional animation will still work on
python2.3. And with the enthought python2.4 release supposedly in the
pipeline, I suspect most who need the faster techniques will be able
to upgrade.
 
 >> P.S. Here's a nit for you: RendererAgg::_get_rgba_face() uses
 >> and returns a std::pair<bool, agg::rgba> instead of a
 >> facepair_t.
 Charlie> Yeah, that's weird that the header and implementation
 Charlie> don't match up, and yet no complaints from the compiler.
One is just a typedef for the other, so while it is not good form,
it's perfectly legal since they both are the same thing. It's
basically like a C macro.
JDH

> Some digging on the internet indicates that that VC++ 6 does not support
> template member functions that are defined outside of the class declarati=
on[1]
> and is badly confused by template memeber functions that don't use the
> parameterized type in their arguments list[2].
>
> I think that your current problem may be solved by moving the definintion=
 of
> _fill_and_stroke() to within the declaration of RendererAgg in
> `backend_agg.h'. I don't have access to a Windows machine at this time, =
so I
> haven't been able to test this hunch myself.
So this approach brought actually got me passed the error, BUT it
walked me straight into an internal compiler error with just a code
returned. I am thinking I will have to start looking into these open
source STL implementations. I'll work on it more tonight and write
back with what I find out. Did I mention my progression through major
programming languages never included C++. I do appreciate and really
need the help.
> P.S. Here's a nit for you: RendererAgg::_get_rgba_face() uses and return=
s a
> std::pair<bool, agg::rgba> instead of a facepair_t.
Yeah, that's weird that the header and implementation don't match up,
and yet no complaints from the compiler.
Thanks,
 Charlie

On 4/13/06, James Evans <jre...@ea...> wrote:
> Charlie,
>
> I have had some experience with both versions of the MSVC++ compiler. Th=
e C++ compiler that
> comes with VS6 is not as good at handling templates, but there are some t=
ricks you can do to
> force the compiler to do what you want.
>
> A lot of times you can achieve this with more formal C++ (remember that t=
he language
> specification lets us take some shortcuts)
>
> Try this on line #362:
>
> this->_fill_and_stroke< agg::ellipse >( path, gc, face );
>
> and do the appropriate similair reeplacement on the other lines. I no lo=
nger have VS6
> installed on any machines that I have access to, so I cannot test this, b=
ut I believe that
> this should work.
Just adding the "this->" didn't change anything, and the compiler
complains about the "< agg::ellipse >", which I don't quite
understand. I think the compiler is complaining more about the first
argument, since it is a template in the declaration. I am seeing more
online that basically says VS6 doesn't handle template very well.
Thanks,
 Charlie

Charlie,
Some digging on the internet indicates that that VC++ 6 does not support 
template member functions that are defined outside of the class declaration[1] 
and is badly confused by template memeber functions that don't use the 
parameterized type in their arguments list[2].
I think that your current problem may be solved by moving the definintion of 
_fill_and_stroke() to within the declaration of RendererAgg in 
`backend_agg.h'. I don't have access to a Windows machine at this time, so I 
haven't been able to test this hunch myself.
Ken
P.S. Here's a nit for you: RendererAgg::_get_rgba_face() uses and returns a 
std::pair<bool, agg::rgba> instead of a facepair_t.
[1] C++ Templates FAQ, Question Four
http://womble.decadentplace.org.uk/c++/template-faq.html#vc6-non-inline-mem
[2] C++ Templates FAQ, Question Five
http://womble.decadentplace.org.uk/c++/template-faq.html#vc6-name-mangling-bug

On Wednesday 12 April 2006 18:13, John Hunter wrote:
> >>>>> "Darren" == Darren Dale <dd...@co...> writes:
>
> Darren> On Wednesday 12 April 2006 17:13, Ryan Krauss wrote:
> >> I am having a font problem with semilogx, usetex=true and
> >> font.family:sans-serif. The xticks are using a serif font.
> >> These lines are enough to recreate the problem on my machine.
> >>
> >> t=arange(0,1,0.01) y=sin(2*pi*t) figure(10) semilogx(t,y)
> >> ylabel('Mag.') xlabel('Time (sec.)') savefig('test.png')
>
> Darren> You might be able to do something like this in
> Darren> LogFormatterMathtext:
>
> Darren> if rcParams['text.usetex']: if not isDecade and
> Darren> self.labelOnlyBase: s = '' elif not isDecade: s =
> Darren> '%d$^\textnormal{%.2f}$'% (b, fx) else: s =
> Darren> '%d^\textnormal{%d}$'% (b, self.nearest_long(fx))
>
> Darren> I dont have time to test it, I'm running out the door.
>
> Hmm, the following is certainly closer than my fool-headed attempt at \rm
>
>
> usetex = rcParams['text.usetex']
> if not isDecade and self.labelOnlyBase: s = ''
> elif not isDecade:
> if usetex: s = r'$\textnormal{%d}^\textnormal{%.2f}$'% (b, fx)
> else: s = '$%d^{%.2f}$'% (b, fx)
> else:
> if usetex: s = r'$\textnormal{%d}^\textnormal{%d}$'% (b,
> self.nearest_long(fx)) else: s = r'$%d^{%d}$'% (b,
> self.nearest_long(fx))
>
>
> but the '1's still look different on my system.
I think I came up with a workable solution, which I committed in svn 2292. I 
also fixed the formatting of the scientific notation label to make it work 
with usetex and sans-serif fonts. The scalarformatter code was my first 
contribution to the mpl codebase, and man is it ugly. I have a hard time 
following the logic anymore. I'll probably refactor it one of these days.
Also, I noticed that the offset functionality isn't working anymore. I think 
it is due to the call to self.viewInterval.span(), the following plot exposes 
the problem:
plot([1,2], array([1, 2])*1e-12+1e-5)
That used to give a plot with properly scaled yaxis, such that the ticks were 
labeled [1,...,2] and the offset text said 'x10^-12 + 1e-5'. Any ideas on 
what has changed?
Darren

Charlie,
I have had some experience with both versions of the MSVC++ compiler. The C++ compiler that
comes with VS6 is not as good at handling templates, but there are some tricks you can do to
force the compiler to do what you want.
A lot of times you can achieve this with more formal C++ (remember that the language
specification lets us take some shortcuts)
Try this on line #362:
this->_fill_and_stroke< agg::ellipse >( path, gc, face );
and do the appropriate similair reeplacement on the other lines. I no longer have VS6
installed on any machines that I have access to, so I cannot test this, but I believe that
this should work.
--James
Date: 2006年4月12日 19:48:43 -0400
From: "Charlie Moad" <cw...@gm...>
To: "matplotlib development list" <mat...@li...>
Subject: [matplotlib-devel] Visual Studio 6 Compiler Error
Hey all,
 So after getting the Visual Studio 7.1/Python 2.4 build working,
I am trying to get Visual Studio 6/Python2.3 working. I could use
some help on a crazy agg/c++ error I am getting. Basically the VS6
compiler is much pickier than VS7.1 I have corrected a few problems,
but I am stuck on this one:
building 'matplotlib.backends._na_backend_agg' extension
C:\Program Files\Microsoft Visual Studio\VC98\BIN\cl.exe /c /nologo
/Ox /MD /W3 /GX /DNDEBUG -Iwin32_static\include -I. -Isrc -Iswig
-Iagg23/include -I. -Iwin32 _static\include -I.
-Iwin32_static\include\freetype2 -I.\freetype2 -Isrc\freetype2
-Iswig\freetype2 -Iagg23/include\freetype2 -I.\freetype2
-Iwin32_static\include\freetype2 -I.\freetype2 -Ic:\Python23\include
-Ic:\Python23\PC /Tpsrc/_na_backend_agg.cpp
Fobuild\temp.win32-2.3\Release\src/_na_backend_agg.obj -DNUMARRAY=3D1
_na_backend_agg.cpp
src/_na_backend_agg.cpp(63) : warning C4800: 'long' : forcing value to
bool 'true' or 'false' (performance warning)
src/_na_backend_agg.cpp(337) : error C2893: Failed to specialize
function template 'void __thiscall RendererAgg::_fill_and_stroke(VS
&,const class GCAgg &,const struct std::pair<bool,struct agg::rgba>
&,bool)'
 With the following template arguments:
 'class agg::path_storage'
src/_na_backend_agg.cpp(362) : error C2893: Failed to specialize
function template 'void __thiscall RendererAgg::_fill_and_stroke(VS
&,const class GCAgg &,const
 struct std::pair<bool,struct agg::rgba> &,bool)'
 With the following template arguments:
 'class agg::ellipse'
src/_na_backend_agg.cpp(411) : error C2893: Failed to specialize
function template 'void __thiscall RendererAgg::_fill_and_stroke(VS
&,const class GCAgg &,const
 struct std::pair<bool,struct agg::rgba> &,bool)'
 With the following template arguments:
 'class agg::path_storage'
src/_na_backend_agg.cpp(1813) : error C2893: Failed to specialize
function template 'void __thiscall RendererAgg::_fill_and_stroke(VS
&,const class GCAgg &,cons
t struct std::pair<bool,struct agg::rgba> &,bool)'
 With the following template arguments:
 'class agg::path_storage'
error: Command ""C:\Program Files\Microsoft Visual
Studio\VC98\BIN\cl.exe" /c /nologo /Ox /MD /W3 /GX /DNDEBUG
-Iwin32_static\include -I. -Isrc -Iswig -Iagg23/include -I.
-Iwin32_static\include -I. -Iwin32_static\include\freetype2
-I.\freetype2 -Isrc\freetype2 -Iswig\freetype2
-Iagg23/include\freetype2 -I.\freetype2
-Iwin32_static\include\freetype2 -I.\freetype2 -Ic:\Python23\include
-Ic:\Python23 \PC /Tpsrc/_na_backend_agg.cpp
/Fobuild\temp.win32-2.3\Release\src/_na_backend_agg.obj -DNUMARRAY=3D1"
failed with exit status 2
Even though it says src/_na_backend_agg.cpp you should be looking at
the _backend_agg.cpp file. My only clue is that class
agg::path_storage does not conform to the VS class, but I don't know
how to fix this.
Thanks,
 Charlie

I am pretty sure that as long as you're getting a refund it doesn't
matter when you file.
You were a grad student all last year right, Darren? You can efile
your federal for free as long as you made less than something like
50ドルk. I did it with TurboTax and it asks you if you want to upgrade
about every 5 minutes. But if you keep saying no, it really is free.
Ryan
On 4/12/06, John Hunter <jdh...@ac...> wrote:
> >>>>> "Darren" =3D=3D Darren Dale <dd...@co...> writes:
>
> Darren> That is my impression as well. I dont have time to look
> Darren> into this more tonight, I have to do my taxes.
>
> Hey, so do I. Do what any reasonable OS developer would due under
> these circumstances:
>
> http://www.irs.gov/formspubs/article/0,,id=3D98155,00.html
>
> :-)
>
> JDH
>

>>>>> "Darren" == Darren Dale <dd...@co...> writes:
 Darren> That is my impression as well. I dont have time to look
 Darren> into this more tonight, I have to do my taxes.
Hey, so do I. Do what any reasonable OS developer would due under
these circumstances:
 http://www.irs.gov/formspubs/article/0,,id=98155,00.html 
:-)
JDH

I can understand not wanting to create a maintenance and support mess.
What do you think about a secret option to add a package (or two)?
I am pretty much o.k. with this as a compromise:
if rcParams['text.usetex']:
 if not isDecade and self.labelOnlyBase: s =3D ''
 elif not isDecade: s =3D '%d$^{%.2f}$'% (b, fx)
 else: s =3D '%d^{%d}$'% (b, self.nearest_long(fx))
The exponents have serifs, but the 10's don't. You have to look
pretty closely to see the serifs on the exponents.
Ryan
On 4/12/06, Darren Dale <dd...@co...> wrote:
> On Wednesday 12 April 2006 6:34 pm, you wrote:
> > Apparently this is not a trivial Latex problem. People have written
> > packages to solve it. I don't know if this solves more problems than
> > it causes, but with helvet+sfmath I get nice results with:
> >
> > ...
> > \usepackage{helvet}
> > \usepackage{sfmath}
> > \begin{document}
> > 10ドル^{-1}$
> > \end{document}
> >
> > sfmath is not a popular package (I don't think), so requiring it as a
> > prerequisite is probably not a good idea. Is it worth re-opening the
> > discussion of a user specified header file for Latex?
>
> I'm hesitant, because people will try to do things that won't work and th=
en
> post here asking for help. I was hoping that usetex was getting to the po=
int
> that I wouldnt have to give it so much attention... but it might be the
> easiest solution to this problem.
>
> > From my poking around and what I have read on the internet and in my
> > Latex book, I don't think we will get san-serif math fonts without a
> > lot of work (or a special package).
>
> That is my impression as well. I dont have time to look into this more
> tonight, I have to do my taxes.
>
> Darren
>

Hey all,
 So after getting the Visual Studio 7.1/Python 2.4 build working,
I am trying to get Visual Studio 6/Python2.3 working. I could use
some help on a crazy agg/c++ error I am getting. Basically the VS6
compiler is much pickier than VS7.1 I have corrected a few problems,
but I am stuck on this one:
building 'matplotlib.backends._na_backend_agg' extension
C:\Program Files\Microsoft Visual Studio\VC98\BIN\cl.exe /c /nologo
/Ox /MD /W3 /GX /DNDEBUG -Iwin32_static\include -I. -Isrc -Iswig
-Iagg23/include -I. -Iwin32 _static\include -I.
-Iwin32_static\include\freetype2 -I.\freetype2 -Isrc\freetype2
-Iswig\freetype2 -Iagg23/include\freetype2 -I.\freetype2
-Iwin32_static\include\freetype2 -I.\freetype2 -Ic:\Python23\include
-Ic:\Python23\PC /Tpsrc/_na_backend_agg.cpp
Fobuild\temp.win32-2.3\Release\src/_na_backend_agg.obj -DNUMARRAY=3D1
_na_backend_agg.cpp
src/_na_backend_agg.cpp(63) : warning C4800: 'long' : forcing value to
bool 'true' or 'false' (performance warning)
src/_na_backend_agg.cpp(337) : error C2893: Failed to specialize
function template 'void __thiscall RendererAgg::_fill_and_stroke(VS
&,const class GCAgg &,const struct std::pair<bool,struct agg::rgba>
&,bool)'
 With the following template arguments:
 'class agg::path_storage'
src/_na_backend_agg.cpp(362) : error C2893: Failed to specialize
function template 'void __thiscall RendererAgg::_fill_and_stroke(VS
&,const class GCAgg &,const
 struct std::pair<bool,struct agg::rgba> &,bool)'
 With the following template arguments:
 'class agg::ellipse'
src/_na_backend_agg.cpp(411) : error C2893: Failed to specialize
function template 'void __thiscall RendererAgg::_fill_and_stroke(VS
&,const class GCAgg &,const
 struct std::pair<bool,struct agg::rgba> &,bool)'
 With the following template arguments:
 'class agg::path_storage'
src/_na_backend_agg.cpp(1813) : error C2893: Failed to specialize
function template 'void __thiscall RendererAgg::_fill_and_stroke(VS
&,const class GCAgg &,cons
t struct std::pair<bool,struct agg::rgba> &,bool)'
 With the following template arguments:
 'class agg::path_storage'
error: Command ""C:\Program Files\Microsoft Visual
Studio\VC98\BIN\cl.exe" /c /nologo /Ox /MD /W3 /GX /DNDEBUG
-Iwin32_static\include -I. -Isrc -Iswig -Iagg23/include -I.
-Iwin32_static\include -I. -Iwin32_static\include\freetype2
-I.\freetype2 -Isrc\freetype2 -Iswig\freetype2
-Iagg23/include\freetype2 -I.\freetype2
-Iwin32_static\include\freetype2 -I.\freetype2 -Ic:\Python23\include
-Ic:\Python23 \PC /Tpsrc/_na_backend_agg.cpp
/Fobuild\temp.win32-2.3\Release\src/_na_backend_agg.obj -DNUMARRAY=3D1"
failed with exit status 2
Even though it says src/_na_backend_agg.cpp you should be looking at
the _backend_agg.cpp file. My only clue is that class
agg::path_storage does not conform to the VS class, but I don't know
how to fix this.
Thanks,
 Charlie

On Wednesday 12 April 2006 6:34 pm, you wrote:
> Apparently this is not a trivial Latex problem. People have written
> packages to solve it. I don't know if this solves more problems than
> it causes, but with helvet+sfmath I get nice results with:
>
> ...
> \usepackage{helvet}
> \usepackage{sfmath}
> \begin{document}
> 10ドル^{-1}$
> \end{document}
>
> sfmath is not a popular package (I don't think), so requiring it as a
> prerequisite is probably not a good idea. Is it worth re-opening the
> discussion of a user specified header file for Latex?
I'm hesitant, because people will try to do things that won't work and then 
post here asking for help. I was hoping that usetex was getting to the point 
that I wouldnt have to give it so much attention... but it might be the 
easiest solution to this problem.
> From my poking around and what I have read on the internet and in my
> Latex book, I don't think we will get san-serif math fonts without a
> lot of work (or a special package).
That is my impression as well. I dont have time to look into this more 
tonight, I have to do my taxes.
Darren

On Wednesday 12 April 2006 6:13 pm, John Hunter wrote:
> >>>>> "Darren" == Darren Dale <dd...@co...> writes:
>
> Darren> On Wednesday 12 April 2006 17:13, Ryan Krauss wrote:
> >> I am having a font problem with semilogx, usetex=true and
> >> font.family:sans-serif. The xticks are using a serif font.
> >> These lines are enough to recreate the problem on my machine.
> >>
> >> t=arange(0,1,0.01) y=sin(2*pi*t) figure(10) semilogx(t,y)
> >> ylabel('Mag.') xlabel('Time (sec.)') savefig('test.png')
>
> Darren> You might be able to do something like this in
> Darren> LogFormatterMathtext:
>
> Darren> if rcParams['text.usetex']: if not isDecade and
> Darren> self.labelOnlyBase: s = '' elif not isDecade: s =
> Darren> '%d$^\textnormal{%.2f}$'% (b, fx) else: s =
> Darren> '%d^\textnormal{%d}$'% (b, self.nearest_long(fx))
>
> Darren> I dont have time to test it, I'm running out the door.
>
> Hmm, the following is certainly closer than my fool-headed attempt at \rm
>
>
> usetex = rcParams['text.usetex']
> if not isDecade and self.labelOnlyBase: s = ''
> elif not isDecade:
> if usetex: s = r'$\textnormal{%d}^\textnormal{%.2f}$'% (b, fx)
> else: s = '$%d^{%.2f}$'% (b, fx)
> else:
> if usetex: s = r'$\textnormal{%d}^\textnormal{%d}$'% (b,
> self.nearest_long(fx)) else: s = r'$%d^{%d}$'% (b,
> self.nearest_long(fx))
>
>
> but the '1's still look different on my system.
>
> Is there any reason you applied the textnormal only to the exponent
> and not the base (other than running out the door)?
My quick attempt only enters math mode to typeset the exponent. Its a useful 
trick for dealing with stoichiometry: SrTiO$_3$